The era of Big Data and data sharing has lead to very large volumes of data becoming available to users across the world. This data is heterogeneous in its modelling, format and quality. Taking full advantage of such data raises many challenges, in particular related to the integration and the understanding of such data. My PhD thesis, started in January 2021, seeks to develop novel methods to help users without advanced IT skills discover a new dataset, by () building an abstract understanding of the data, as consisting of records and collections, () interpreting or classifying the data based on users' interests, and leveraging Information Extraction and Natural Language tools.